复杂地表条件下低分数据参考的中分辨率遥感数据辐射归一化方法

The inconsistency is commonly existing, and thus limits the potential realizing of the amounts of middle resolution remote sensing data. Utilizing the coarser resolution data as reference to normalize multi-source data is a high-functional way to eliminate the inconsistency. This project is going to solve some key issues about the normalization over complex land surface. Aiming at the two difficulties, the high quantity sample set acquisition and the non-linear radiometric transforming relationship modeling. An accurate sample set optimization method will be researched to guarantee both the quantity and quality of the sample set, in which the low-quality sample will be removed and the scale effect correction will be taken to the others. Meanwhile the non-parametric regression method will be introduced to accurately model the complex non-linear transforming relation among multi-sensor data. Moreover, the variational model with spatial-temporal constraint will be taken to improve the overall accuracy of multiple normalizing results. Through the above comprehensive researches about multi-source data normalization, it is hoped that the inconsistency among different data can be efficiently eliminated, to make the middle resolution data be convenient to be used for kinds of applications, such as urban monitoring, environment protection and so on.

辐射不一致问题普遍存在于多源遥感数据之间，制约了海量数据的集成利用，以低分数据作为辐射参考基准的辐射归一化是解决这一问题的有效途径。项目针对复杂地表条件下高精度建模样本难以获取、非线性辐射转换关系复杂难以刻画的问题,研究复杂地表下的样本集优化策略：通过自适应阈值调整剔除低同质数据，通过神经网络尺度效应校正模型，校正待归一化数据尺度上推结果的尺度效应；引入统计学方法和泛函理论，利用非参数回归方法挖掘数据自身信息，实现数据的高精度非线性辐射转换；通过时空关联约束变分模型提高多源数据归一化结果的整体一致性。研究成果可实现复杂地表条件下时空多源数据的高精度辐射归一化，为应用提供可靠、可比、统一的数据，提升海量遥感数据应用潜力，具有重要的研究价值。

辐射不一致问题普遍存在于多源遥感数据之间，制约了海量多源数据的集成利用。项目采用以低分数据作为辐射参考基准的归一化方法，针对复杂地表条件下难以获得高精度建模样本点、非线性辐射转换关系复杂难以刻画的问题, 在“样本获取—模型构建—整体求解”的主线方向以及辐射归一化在生成高时空密度遥感数据中的应用等方面开展了研究，并在样本分布及归一化应用方面开展了扩展研究。提出了基于修正变异指数的纯净象元提取指标，研究了样本分布对归一化的影响分析并提出了对应的改进的归一化方法；建立了地理加权非参数回归模型；建立了基于多重约束的多个数据整体归一化模型；开展了辐射归一化与时空融合相结合的高时空分辨率数据生成，以及基于辐射归一化的云南地区植被净初级生产力时空变化分析。研究成果实现了基于低分数据参考的时空多源中分辨率数据的高精度辐射归一化，为应用提供可靠、可比、统一的数据，具有重要的理论意义和应用价值。.按照既定的研究计划，项目圆满地完成了课题设定的研究计划，实现了预期目标，并在研究的深度和广度上都比既定计划都有了进一步的扩展。在项目资助下，项目组发表科研论文5篇，其中4篇SCI检索，待发表英文期刊论文约4篇；在人才培养方面项目组培养在读硕士生3名。